They spent £16 million on Witney STW. Did it actually work?

The River Windrush is a chalk stream — rare, clear, cold, the kind of watercourse that shows up in old landscape paintings and EA enforcement records in roughly equal measure. It runs through Witney, West Oxfordshire, downstream of a sewage treatment works that has been on the Environment Agency's radar for years. So when Thames Water announced a £16 million upgrade to Witney STW in October 2025, the question worth asking was: does the data back it up?

The Windrush Action Group has been documenting discharge events on this stretch of river for years. Local press — including the Oxford Mail and Witney Gazette — covered the upgrade announcement, and campaigners were cautious: Thames Water has a habit of announcing improvements that take a long time to show up in the numbers. Five months on, here's what WaterWatch's analysis actually finds.

What they actually built — and why it took this long

The Windrush isn't a new story. The Windrush Action Group has been documenting sewage discharges for years, and the river regularly appears in EA enforcement records and local news coverage. It's one of those sites where the long-term spill history is genuinely difficult to look at without feeling slightly depressed.

The £16 million programme expanded Witney STW's treatment capacity by two-thirds — from 240 to 399 litres per second — specifically to reduce the number of times the works has to bypass treatment during heavy rainfall. Thames Water have committed to meeting all government storm overflow targets at CATM.3134 by 2040–2045. (Yes, that's almost twenty years away. The Environment Act passed in 2021. Make of that what you will.)

The upgrade completed in autumn 2025. The EDM sensor — the little monitor that tells WaterWatch when sewage is being discharged — started recording a new baseline from November onwards. That's what we're looking at here.

The awkward problem with "before and after" comparisons

The obvious thing to do when an upgrade finishes is to count the spills before and compare them to the spills after. Thames Water had 120 recorded discharge episodes at CATM.3134 across the two years before the upgrade. In the five months since? Thirty-two. That looks impressive on the surface — until you notice that you're comparing 731 days against 163 days, and conclude that most of the difference is just… time.

Simple spill counts are also sensitive to weather. If the year before the upgrade was particularly wet and the year after was unusually dry, any apparent improvement might have nothing to do with the works at all — it could just be luck of the calendar. Raw counts are a good pub statistic. They're a bad way to evaluate infrastructure.

So WaterWatch doesn't use raw counts. Instead, I calculate a daily "environmental pressure score" for each site — combining 24-hour rainfall, antecedent soil wetness (how saturated the ground was going into any given day), river level anomaly, and — new in v1.6 — groundwater depth. Then I count how many spills occurred per 100 units of that pressure, in the periods either side of the upgrade. Weather-adjusted. Apples to apples.

If you want the full technical picture of how this works — the maths, the data sources, the edge cases — the methodology post covers it in detail. The short version for here: we're comparing like-for-like weather conditions, so what's left should reflect the upgrade.

What the data actually shows

The spill rate fell from 1.63 to 1.26 per 100 units of environmental pressure — a 23% reduction. That's the headline figure. But there's a detail buried in the numbers that makes this finding more interesting than it first appears.

The after period (November 2025 to April 2026) was a winter period. The average daily environmental pressure was 15.6 units per day, versus 10.1 in the two-year window before the upgrade. In other words: the site was, on average, under harder conditions after the upgrade than before it. The fact that the spill rate still fell by 23% under those conditions suggests the new capacity is doing real work — not just sitting idle while the weather happened to cooperate.

At medium environmental pressure (rainfall stress between 3 and 6 on the index), the site recorded zero discharge episodes in the post-upgrade period, compared to a rate of 0.0017 per pressure unit before. That's the pressure range where you'd expect the extra 159 litres per second of capacity to make a noticeable difference — and it appears to have done exactly that. At high pressure (storm events), the rate dropped from 0.0181 to 0.0129. Progress, but the tough conditions are still producing spills.

The groundwater problem

The Windrush catchment sits on Cotswold limestone — porous oolite that holds enormous quantities of water and releases it slowly into river gravels and, inevitably, into the sewer network. Groundwater inundation is a specific failure mode for combined sewers in this part of Oxfordshire: during prolonged wet winters, rising water tables push groundwater directly into pipe joints, increasing the base flow reaching the treatment works even on days with no rainfall whatsoever.

This matters for the analysis because groundwater inundation creates a category of sewer stress that neither rainfall data nor river levels fully capture. A site can show elevated spill frequency in January not because it rained heavily that week, but because the water table has been rising since October. WaterWatch's v1.6 analysis now incorporates groundwater depth from the nearest Environment Agency monitoring borehole — adding it as a fourth input to the daily pressure score, scaled and smoothed with a carry-forward to reflect how slowly groundwater levels change.

For CATM.3134 specifically, the groundwater contribution is material. The Windrush valley boreholes showed elevated water table readings through the winter of 2025–26 — conditions that would have inflated the raw pressure score had they not been explicitly accounted for. Including groundwater makes the after-window pressure calculation more accurate, and means the 23% rate reduction is, if anything, a conservative estimate of the upgrade's effect: the sewer was under more combined stress than rainfall alone suggested, yet still spilled less.

Thames Water cited groundwater inundation as a contributing factor in the site's historical spill pattern during the planning consultation for this upgrade. The v1.6 analysis now measures that contribution directly rather than leaving it as an unexplained residual.

This is a strong indication, not a verdict

I want to be honest about what five months of data can and cannot tell us. The confidence rating on this analysis is NORMAL— not HIGH. That's because 163 days is a short post-upgrade window, and short windows can mislead. A string of dry months would produce good numbers regardless of any upgrade. A particularly brutal winter storm season could produce bad ones even with a doubled-capacity works.

There are other factors WaterWatch cannot fully control for. Population changes in Witney increase base load gradually. The EDM sensor itself can record brief offline periods or false starts that inflate spill counts — EDM monitoring is essential infrastructure, but it is not infallible, and understanding its limitations is part of reading this data honestly. While groundwater is now included in the pressure model, the borehole network is sparse — one station covers a wide catchment, and local variation in water table depth near the works may differ from the monitored point.

What the data does show, fairly clearly, is that the site is performing differently since the upgrade — and the direction of travel is the right one. If this trend holds through 2026 and 2027, with a full range of seasonal and storm conditions in the post-upgrade window, the confidence rating will move up.

Where next for Witney

The £16 million capacity upgrade addresses one half of the problem: when sewage arrives at the gates of Witney STW faster than the works can treat it, water now has 159 extra litres per second of headroom before it overflows. But there is a second, quieter problem that extra capacity alone cannot fix — groundwater that is already inside the pipes before it ever reaches the works.

Thames Water has published a Groundwater Impacted System Management Plan for Witney — a separate programme specifically targeting infiltration reduction. Where the capacity upgrade is about processing more flow, the management plan is about stopping groundwater from entering the sewer network in the first place: CCTV surveys to locate cracked and leaking pipe sections, lining and sealing of the worst-affected sewers, manhole rehabilitation, and systematic monitoring of infiltration volumes at key points in the network.

This distinction matters. A sewer running in a high water-table area like the Windrush valley can receive enormous volumes of clean groundwater through pipe joints and defects — water that occupies treatment capacity, dilutes flow, and contributes to spills without ever having been near a property. Lining a pipe is unglamorous work; it doesn't get press releases the way a new tank does. But for a chalk-stream catchment like this one, it may ultimately move the numbers more than any single infrastructure upgrade.

The management plan represents Thames Water's acknowledgement that the Witney system has a structural groundwater problem that requires active management, not just capacity headroom. WaterWatch will track both workstreams — if infiltration volumes fall significantly in coming years, that should show up in the pressure scores independently of any further capital works.

How we ran this analysis

Two comparison windows were used: Sep 2023–Aug 2025 (before the upgrade, 731 days) and Nov 2025–Apr 2026 (after the upgrade, 163 days). A ±30-day exclusion window either side of October 2025 removes any spills that happened during construction disruption or the settling-in period — otherwise the data around the upgrade date would be noisy in both directions.

Each day is assigned an environmental pressure score using four inputs: 24-hour rainfall from the nearest EA rain gauge, antecedent wetness index (AWI) calculated using Kohler & Linsley's exponential decay model (k = 0.85 — yesterday's rain counts for 85% of today's contribution to soil saturation), river level anomaly against seasonal norms, and groundwater depth from the nearest EA monitoring borehole (carry-forward smoothed to reflect the slow response of water table levels). Spill rate = total spills ÷ total pressure × 100. Data quality was good in both windows: the assigned rain station had no significant coverage gaps.

Analysis: WaterWatch improvement analysis tool (v1.6) · Permit: CATM.3134 · October 2025 upgrade · Analysed: 2026-04-12 · Methodology: How we analyse infrastructure upgrades